/// <summary>
/// show detected features within the given image
/// </summary>
/// <param name="img"></param>
public void ShowFeatures(classimage_mono img, int feature_type)
{
show_ellipses = false;
switch (feature_type)
{
case MonoSLAM.DISPLAY_FEATURES:
{
monoslaminterface.ShowFeatures(img);
break;
}
case MonoSLAM.DISPLAY_ELLIPSES:
{
show_ellipses = true;
monoslaminterface.ShowFeatures(img);
break;
}
case MonoSLAM.DISPLAY_PROBABILITIES:
{
monoslaminterface.showProbabilities(0, img);
//monoslaminterface.ShowTriangles(img);
//monoslaminterface.ShowLineHistory(img);
//monoslaminterface.ShowMesh(img);
break;
}
case MonoSLAM.DISPLAY_MAP:
{
monoslaminterface.ShowOverheadView(img, 2.5f, 2.5f);
break;
}
}
}
/// <summary>
/// load a new image from the camera
/// </summary>
/// <param name="imageptr">camera image</param>
public virtual void load_new_image(classimage_mono imageptr, classimage imageptr_colour, classimage_mono outputimageptr, bool show_ellipses)
{
image = imageptr;
image_colour = imageptr_colour;
outputimage = outputimageptr;
this.show_ellipses = show_ellipses;
}
/// <summary>
/// Make the first measurement of a feature.
/// </summary>
/// <param name="z">The location of the feature to measure</param>
/// <param name="id">The Identifier to fill with the feature measurements</param>
/// <param name="f_m_m">The feature measurement model to use for this partially-initialised feature</param>
/// <returns></returns>
public virtual bool make_initial_measurement_of_feature(Vector z,
ref classimage_mono id,
Partially_Initialised_Feature_Measurement_Model f_m_m,
Vector patch_colour)
{
return(false);
}
/// <summary>
/// Constructor for known features. The different number of
/// arguments differentiates it from the constructor for partially-initialised
/// features
/// </summary>
/// <param name="id">reference to the feature identifier</param>
/// <param name="?"></param>
public Feature(classimage_mono id, uint lab, uint list_pos,
Scene_Single scene, Vector y_known,
Vector xp_o,
Feature_Measurement_Model f_m_m, uint k_f_l)
{
feature_measurement_model = f_m_m;
feature_constructor_bookeeping();
identifier = id;
label = lab;
position_in_list = list_pos; // Position of new feature in list
// Save the vehicle position where this feature was acquired
xp_orig = new Vector(xp_o);
// Straighforward initialisation of state and covariances
y = y_known;
Pxy = new MatrixFixed(scene.get_motion_model().STATE_SIZE, feature_measurement_model.FEATURE_STATE_SIZE);
Pxy.Fill(0.0f);
Pyy = new MatrixFixed(feature_measurement_model.FEATURE_STATE_SIZE, feature_measurement_model.FEATURE_STATE_SIZE);
Pyy.Fill(0.0f);
int i = 0;
MatrixFixed newPyjyi_to_store;
foreach (Feature it in scene.get_feature_list_noconst())
{
if (i < position_in_list)
{
newPyjyi_to_store = new MatrixFixed(
it.get_feature_measurement_model().FEATURE_STATE_SIZE,
feature_measurement_model.FEATURE_STATE_SIZE);
//add to the list
matrix_block_list.Add(newPyjyi_to_store);
}
i++;
}
known_feature_label = (int)k_f_l;
if (feature_measurement_model.fully_initialised_flag)
{
partially_initialised_feature_measurement_model = null;
fully_initialised_feature_measurement_model =
(Fully_Initialised_Feature_Measurement_Model)feature_measurement_model;
}
else
{
fully_initialised_feature_measurement_model = null;
partially_initialised_feature_measurement_model =
(Partially_Initialised_Feature_Measurement_Model)feature_measurement_model;
}
}
/// <summary>
/// display the line history as an image
/// </summary>
/// <param name="img"></param>
public void show(classimage_mono img)
{
for (int x = 0; x < img.width; x++)
{
int xx = x * history / img.width;
for (int y = 0; y < img.height; y++)
{
int yy = y * no_of_points / img.height;
img.image[x, y] = pixel_intensity[yy, xx];
}
}
}
/// <summary>
/// Constructor
/// </summary>
/// <param name="image">The image to search</param>
/// <param name="patch">The image patch to search for</param>
/// <param name="BOXSIZE">The size of the image patch to use</param>
public SearchMultipleOverlappingEllipses(classimage_mono image, classimage_mono patch, uint BOXSIZE)
{
m_image = image;
m_patch = patch;
m_boxsize = BOXSIZE;
// Rather than working out an overall bounding box, we'll be slightly
// lazy and make an array for correlation results which is the same size
// as the image
// Pixels in this array in raster order
correlation_image = new float[m_image.width, m_image.height];
}
//**************************Write Image Patch to Disk**************************
/// <summary>
/// Save the currently-selected patch to a file.
/// </summary>
/// <param name="name"></param>
public void write_patch(String name)
{
classimage_mono hip = new classimage_mono();
hip.createImage((int)Camera_Constants.BOXSIZE, (int)Camera_Constants.BOXSIZE);
if (location_selected_flag)
{
// Copy the selected patch to the save space patch
copy_into_patch(image, hip, uu, vv);
hip.SaveAsBitmapMono(name);
}
}
/// <summary>
/// Little service function to copy image region centred at uu, vv into patch
/// </summary>
/// <param name="im">the image to extract the patch from</param>
/// <param name="patch">the patch to be returned</param>
/// <param name="uu">x coordinate of the centre of the patch within the image</param>
/// <param name="vv">y coordinate of the centre of the patch within the image</param>
public void copy_into_patch(classimage_mono im, classimage_mono patch,
uint uu, uint vv)
{
for (uint r = 0; r < Camera_Constants.BOXSIZE; r++)
{
for (uint c = 0; c < Camera_Constants.BOXSIZE; c++)
{
int x = (int)(c + uu - (Camera_Constants.BOXSIZE - 1) / 2);
int y = (int)(r + vv - (Camera_Constants.BOXSIZE - 1) / 2);
patch.image[c, r] = im.image[x, y];
}
}
}
/// <summary>
/// update the line history
/// </summary>
/// <param name="img"></param>
public void update(classimage_mono img)
{
int i;
Vector position1 = feature1.get_z();
Vector position2 = feature2.get_z();
// store the feature positions
for (i = 0; i < 3; i++)
{
feature_position[curr_index, 0][i] = feature1.get_y()[i];
feature_position[curr_index, 1][i] = feature2.get_y()[i];
}
// distances between features in image coordinates
float dist_u = position1[0] - position2[0];
float dist_v = position1[1] - position2[1];
marked_for_deletion = false;
for (i = 0; i < no_of_points; i++)
{
int x = (int)(position1[0] + (i * dist_u / no_of_points));
if ((x > 0) && (x < img.width))
{
int y = (int)(position1[1] + (i * dist_v / no_of_points));
if ((y > 0) && (y < img.height))
{
pixel_intensity[i, curr_index] = img.image[x, y];
}
else
{
marked_for_deletion = true;
}
}
else
{
marked_for_deletion = true;
}
}
curr_index++;
if (curr_index >= history)
{
curr_index = 0;
}
if (hits < 254)
{
hits++;
}
}
/// <summary>
/// Initialise a known feature. In this case it is assumed that a known feature
/// is an image patch to be loaded from a file <code>known_patch?.bmp</code>.
/// </summary>
/// <param name="fmm"></param>
/// <param name="v"></param>
/// <param name="known_feature_label"></param>
/// <returns></returns>
public override classimage_mono initialise_known_feature(Feature_Measurement_Model fmm,
Vector v, uint known_feature_label, String path)
{
String name = Camera_Constants.known_point_patch_stem + known_feature_label + ".bmp";
classimage_mono patch = new classimage_mono();
if (!patch.loadFromBitmapMono(path + name, (int)Camera_Constants.BOXSIZE, (int)Camera_Constants.BOXSIZE))
{
patch = null;
}
return(patch);
}
/// <summary>
/// Make measurements of a feature which is represented by a set of particles.
/// This is typically a partially-initialised feature (see
/// Partially_Initialised_Feature_Measurement_Model), where the particles represent
/// the probability distribution in the direction of the free parameters.
/// </summary>
/// <param name="id">The Identified for this feature (in this case this will be an image patch)</param>
/// <param name="particle_vector">The vector of particles. The covariance of each particle is used to define the region over which the feature is matched and measured.</param>
public void measure_feature_with_multiple_priors(classimage_mono patch, ArrayList particle_vector)
{
SearchMultipleOverlappingEllipses ellipse_search =
new SearchMultipleOverlappingEllipses(image, patch, Camera_Constants.BOXSIZE);
SearchDatum e;
foreach (Particle part in particle_vector)
{
ellipse_search.add_ellipse(part.get_SInv(), part.get_h());
//if (Camera_Constants.DEBUGDUMP) cout << "Particle at " << part->get_lambda()
// << " h " << part->get_h()
// << " SInv " << part->get_SInv() << endl;
}
// cout << "MFWMP timer before search_multiple: " << timerlocal << endl;
ellipse_search.search();
// cout << "MFWMP timer after search_multiple: " << timerlocal << endl;
// Turn results into matrix forms
int i = 0;
foreach (Particle it in particle_vector)
{
e = (SearchDatum)ellipse_search.m_searchdata[i];
if (e.result_flag)
{
// Save the measurement location back into the particle
Vector z_local = new Vector(2);
z_local[0] = e.result_u;
z_local[1] = e.result_v;
it.set_z(z_local);
it.set_successful_measurement_flag(true);
}
else
{
it.set_successful_measurement_flag(false);
}
i++;
}
// cout << "MFWMP timer end: " << timerlocal << endl;
}
/// <summary>
/// Initialise a known feature, in this case by loading an image file. The name of
/// the file is read from the Settings Section passed to this function, with the
/// entry Identifier.
/// </summary>
/// <param name="fmm"></param>
/// <param name="v"></param>
/// <param name="section"></param>
/// <returns></returns>
public override classimage_mono initialise_known_feature(Feature_Measurement_Model fmm,
Vector v, Settings.Section section,
String path)
{
ArrayList values = section.get_entry("Identifier");
String name = (String)values[0];
//cout << "Reading patch " << name << endl;
classimage_mono patch = new classimage_mono();
if (!(patch.loadFromBitmapMono(path + name, (int)Camera_Constants.BOXSIZE, (int)Camera_Constants.BOXSIZE)))
{
patch = null;
}
return(patch);
}
/// <summary>
/// Make the initial measurement of the currently-selected feature. This fills in
/// the location and the identifier (a copy of the image patch) for the current
/// feature. The feature location is set using set_image_selection_automatically()
/// or manually by the user using set_image_selection(). This function just calls
/// partially_initialise_point_feature() to fill in the measurement and identifier.
/// </summary>
/// <param name="z">The measurement vector to be filled in.</param>
/// <param name="id_ptr">The identifier to be filled in.</param>
/// <param name="m"></param>
/// <returns>true on success, or <code>false</code> on failure (e.g. no patch is currently selected).</returns>
public override bool make_initial_measurement_of_feature(Vector z, ref classimage_mono patch, Partially_Initialised_Feature_Measurement_Model m, Vector patch_colour)
{
patch = partially_initialise_point_feature(z);
for (int c = 0; c < 3; c++)
{
patch_colour[c] = image_colour.image[uu, vv, c];
}
if (patch != null)
{
return(true);
}
else
{
return(false);
}
}
/// <summary>
/// show the mesh in the given image
/// </summary>
/// <param name="img"></param>
public void Show(classimage_mono img)
{
for (int i = 0; i < features.Count; i++)
{
Feature feat = (Feature)features[i];
if (featureVisible(feat))
{
for (int j = 0; j < feat.triangles.Count; j++)
{
model_triangle tri = (model_triangle)feat.triangles[j];
if (tri.isVisible())
{
tri.Show(img);
}
}
}
}
}
/// <summary>
/// Make a measurement of a feature. This function calls elliptical_search() to
/// find the best match within three standard deviations of the predicted location.
/// </summary>
/// <param name="patch">The identifier for this feature (in this case an image patch)</param>
/// <param name="z">The best image location match for the feature, to be filled in by this function</param>
/// <param name="h">The expected image location</param>
/// <param name="S">The expected location covariance, used to specify the search region.</param>
/// <returns></returns>
public override bool measure_feature(classimage_mono patch, ref Vector z, Vector vz, Vector h, MatrixFixed S, Random rnd)
{
Cholesky S_cholesky = new Cholesky(S);
MatrixFixed Sinv = S_cholesky.Inverse();
uint u_found = 0, v_found = 0;
if (SceneLib.elliptical_search(image, patch,
h, Sinv, ref u_found, ref v_found, vz, Camera_Constants.BOXSIZE, outputimage, show_ellipses, calibrating, rnd) != true)
{
// Feature not successfully matched
return(false);
}
z.Put(0, (float)u_found);
z.Put(1, (float)v_found);
return(true);
}
public static void find_best_patch_inside_region_test(
classimage_mono image,
ref uint ubest, ref uint vbest, ref float evbest,
uint BOXSIZE, uint ustart, uint vstart,
uint ufinish, uint vfinish)
{
long corr;
long corrmax = MIN_PATCH_DIFFERENCE * BOXSIZE * BOXSIZE / 2;
int tx, ty, bx, by;
for (int x = (int)ustart; x < ufinish; x++)
{
for (int y = (int)vstart; y < vfinish; y++)
{
tx = (int)(x - (BOXSIZE - 1) / 2);
ty = (int)(y - (BOXSIZE - 1) / 2);
bx = (int)(tx + BOXSIZE);
by = (int)(ty + BOXSIZE);
long leftval = image.getIntegral(tx, ty, x, by);
long rightval = image.getIntegral(x, ty, bx, by);
long topval = image.getIntegral(tx, ty, bx, y);
long bottomval = image.getIntegral(tx, y, bx, by);
corr = Math.Abs(leftval - rightval) +
Math.Abs(topval - bottomval) +
Math.Abs(leftval - topval) +
Math.Abs(rightval - topval) +
Math.Abs(leftval - bottomval) +
Math.Abs(rightval - bottomval);
if (corr > corrmax)
{
corrmax = corr;
ubest = (uint)x;
vbest = (uint)y;
evbest = corr;
}
}
}
}
public void Show(classimage_mono img)
{
int i, j;
int[,] screen_position = new int[3, 2];
for (i = 0; i < 3; i++)
{
Vector screen_pos = ((model_vertex)vertices[i]).vertex_feature.get_h();
screen_position[i, 0] = (int)screen_pos[0];
screen_position[i, 1] = (int)screen_pos[1];
}
for (i = 0; i < 3; i++)
{
for (j = 0; j < 3; j++)
{
if (i != j)
{
img.drawLine(screen_position[i, 0], screen_position[i, 1],
screen_position[j, 0], screen_position[j, 1], 0);
}
}
}
}
/// <summary>
/// Creates a new ImageMonoExtraData to represent the currently-selected image
/// patch, and also returns its location in the parameter z. The current
/// image patch is set using set_image_selection_automatically() or manually by the
/// user using set_image_selection().
/// </summary>
/// <param name="z">The measurement vector to be filled in</param>
/// <returns>The classimage holding this image patch information (created using new, so must be deleted elsewhere), or null if no patch is currently selected.</returns>
public classimage_mono partially_initialise_point_feature(Vector z)
{
if (location_selected_flag) // Patch selected
{
// Go and initialise it in scene + 3D
classimage_mono hip = new classimage_mono();
hip.createImage((int)Camera_Constants.BOXSIZE, (int)Camera_Constants.BOXSIZE);
// Final save of fixated image patch
copy_into_patch(image, hip, uu, vv);
// And set measurement
z.Put(0, uu);
z.Put(1, vv);
// return the patch
return(hip);
}
else
{
// No patch selected
return(null);
}
}
/// <summary>
/// Make a measurement of a feature.
/// </summary>
/// <param name="id">The identifier for this feature</param>
/// <param name="z">The measurement of the state, to be filled in by this function</param>
/// <param name="h">The expected measurement</param>
/// <param name="S">The measurement covariance.</param>
/// <returns></returns>
public virtual bool measure_feature(classimage_mono id, ref Vector z, Vector vz, Vector h, MatrixFixed S, Random rnd)
{
return(false);
}
/// <summary>
/// one update
/// </summary>
/// <param name="img"></param>
public void GoOneStep(classimage_mono img, classimage img_colour, classimage_mono output_img)
{
float delta_t;
MAXIMUM_ANGLE_DIFFERENCE = 3.1415927f * (field_of_vision / 2) / 180.0f;
//set calibration parameters
if (calibrating)
{
monoslaminterface.set_camera_calibration(9e-06f, lens_distortion, lens_distortion, centre_of__distortion_x, centre_of__distortion_y, 1);
}
//load in the current raw image
Robot r = monoslaminterface.GetRobot();
r.calibrating = (!enable_mapping) | calibrating;
r.load_new_image(img, img_colour, output_img, show_ellipses);
stopwatch.Stop();
//get the elapsed time in seconds
delta_t = stopwatch.ElapsedMilliseconds / 1000.0f;
//if using the simulation set the frame rate at a fixed value
if (simulation_mode)
{
delta_t = 1.0f / 20.0f;
}
if (delta_t > 0)
{
frames_per_second = (int)((1.0f / delta_t) * 10) / 10.0f;
//report if its taking a long time
frame_rate_warning = false;
if (delta_t > 0.2)
{
frame_rate_warning = true;
Debug.WriteLine("Time between frames (sec): " + Convert.ToString(delta_t));
}
//update the state of the system
monoslaminterface.GoOneStep(delta_t, enable_mapping);
}
stopwatch.Reset();
stopwatch.Start();
speed = (int)(monoslaminterface.Speed() * 100) / 100.0f;
if ((enable_mapping) && (monoslaminterface.number_of_matched_features < 2))
{
enable_mapping = false;
init();
}
if ((!enable_mapping) && (monoslaminterface.number_of_matched_features >= 3) && (!calibrating))
{
registered_seconds += delta_t;
if (registered_seconds > 1)
{
Debug.WriteLine("Ready to go");
enable_mapping = true;
registered_seconds = 0;
}
}
else
{
if (registered_seconds > 0)
{
Debug.WriteLine("Waiting for initial registration");
}
registered_seconds = 0;
}
}
/// <summary>
/// Normalise sum-of-squared-difference between two patches. Assumes that the images
/// are continuous memory blocks in raster order. The two pointer arguments return
/// the standard deviations of the patches: if either of these are low, the results
/// can be misleading.
/// </summary>
/// <param name="x0">Start x-co-ordinate of patch in first image</param>
/// <param name="y0">End x-co-ordinate of patch in first image</param>
/// <param name="x0lim">End x-co-ordinate of patch in first image</param>
/// <param name="y0lim">End y-co-ordinate of patch in first image</param>
/// <param name="x1">Start x-co-ordinate of patch in second image</param>
/// <param name="y1">Start y-co-ordinate of patch in second image</param>
/// <param name="p0">First image</param>
/// <param name="p1">Second image</param>
/// <param name="sd0ptr">Standard deviation of patch from first image</param>
/// <param name="sd1ptr">Standard deviation of patch from second image</param>
/// <returns></returns>
public static float correlate2_warning(
int x0, int y0,
int x0lim, int y0lim,
int x1, int y1,
ref classimage_mono p0, ref classimage_mono p1,
ref float sd0ptr, ref float sd1ptr)
{
// Make these int rather than unsigned int for speed
int patchwidth = x0lim - x0;
int p0skip = p0.width - patchwidth;
int p1skip = p1.width - patchwidth;
int Sg0 = 0, Sg1 = 0, Sg0g1 = 0, Sg0sq = 0, Sg1sq = 0;
float n = (x0lim - x0) * (y0lim - y0); // to hold total no of pixels
float varg0 = 0.0f, varg1 = 0.0f, sigmag0 = 0.0f, sigmag1 = 0.0f, g0bar = 0.0f, g1bar = 0.0f;
// variances, standard deviations, means
float Sg0doub = 0.0f, Sg1doub = 0.0f, Sg0g1doub = 0.0f, Sg0sqdoub = 0.0f, Sg1sqdoub = 0.0f;
float C = 0.0f; // to hold the final result
float k = 0.0f; // to hold an intermediate result
// at the moment the far right and bottom pixels aren't included
int v0, v1;
for (int y0copy = y0lim - 1; y0copy >= 0; y0copy--)
{
int yy0 = y0 + y0copy;
int yy1 = y1 + y0copy;
for (int x0copy = x0lim - 1; x0copy >= 0; x0copy--)
{
v0 = p0.image[x0 + x0copy, yy0];
v1 = p1.image[x1 + x0copy, yy1];
Sg0 += v0;
Sg1 += v1;
Sg0g1 += v0 * v1;
Sg0sq += v0 * v0;
Sg1sq += v1 * v1;
}
}
Sg0doub = Sg0;
Sg1doub = Sg1;
Sg0g1doub = Sg0g1;
Sg0sqdoub = Sg0sq;
Sg1sqdoub = Sg1sq;
g0bar = Sg0doub / n;
g1bar = Sg1doub / n;
varg0 = Sg0sqdoub / n - (g0bar * g0bar);
varg1 = Sg1sqdoub / n - (g1bar * g1bar);
sigmag0 = (float)Math.Sqrt(varg0);
sigmag1 = (float)Math.Sqrt(varg1);
sd0ptr = (float)sigmag0;
sd1ptr = (float)sigmag1;
if (sigmag0 == 0.0f) // special checks for this algorithm
{ // to avoid division by zero
if (sigmag1 == 0.0f)
{
return(0.0f);
}
else
{
return(1.0f);
}
}
if (sigmag1 == 0.0f)
{
return(1.0f);
}
k = g0bar / sigmag0 - g1bar / sigmag1;
C = Sg0sqdoub / varg0 + Sg1sqdoub / varg1 + n * (k * k)
- Sg0g1doub * 2.0f / (sigmag0 * sigmag1)
- Sg0doub * 2.0f * k / sigmag0 + Sg1doub * 2.0f * k / sigmag1;
return(C / n); // returns mean square no of s.d. from mean of pixels
}
/// <summary>
/// Function to scan over (a window in an) image and find the best patch by the Shi
/// and Tomasi criterion.
/// Method: as described in notes from 1/7/97. Form sums in an incremental
/// way to speed things up.
/// </summary>
/// <param name="image">The image to scan</param>
/// <param name="ubest">The x-co-ordinate of the best patch</param>
/// <param name="vbest">The y-co-ordinate of the best patch</param>
/// <param name="evbest">The smallest eigenvalue of the best patch (larger is better)</param>
/// <param name="BOXSIZE">The size of the patch to use</param>
/// <param name="ustart">The x-co-ordinate of the start of the search window</param>
/// <param name="vstart">The y-co-ordinate of the start of the search window</param>
/// <param name="ufinish">The x-co-ordinate of the end of the search window</param>
/// <param name="vfinish">The v-co-ordinate of the end of the search window</param>
public static void find_best_patch_inside_region(
classimage_mono image,
ref uint ubest, ref uint vbest, ref float evbest,
uint BOXSIZE, uint ustart, uint vstart,
uint ufinish, uint vfinish)
{
// Check that these limits aren't too close to the image edges.
// Note that we can't use the edge pixels because we can't form
// gradients here.
if (ustart < (BOXSIZE - 1) / 2 + 1)
{
ustart = (BOXSIZE - 1) / 2 + 1;
}
if (ufinish > image.width - (BOXSIZE - 1) / 2 - 1)
{
ufinish = (uint)(image.width - (BOXSIZE - 1) / 2 - 1);
}
if (vstart < (BOXSIZE - 1) / 2 + 1)
{
vstart = (BOXSIZE - 1) / 2 + 1;
}
if (vfinish > image.height - (BOXSIZE - 1) / 2 - 1)
{
vfinish = (uint)(image.height - (BOXSIZE - 1) / 2 - 1);
}
// Is there anything left to search? If not, set the score to zero and return.
if ((vstart >= vfinish) || (ustart >= ufinish))
{
ubest = ustart;
vbest = vstart;
evbest = 0;
return;
}
// Calculate the width we need to find gradients in.
uint calc_width = ufinish - ustart + BOXSIZE - 1;
// Arrays which store column sums of height BOXSIZE
float[] CSgxsq = new float[calc_width];
float[] CSgysq = new float[calc_width];
float[] CSgxgy = new float[calc_width];
// For the final sums at each position (u, v)
float TSgxsq = 0.0f, TSgysq = 0.0f, TSgxgy = 0.0f;
float gx, gy;
uint u = ustart, v = vstart;
float eval1 = 0, eval2 = 0;
// Initial stage: fill these sums for the first horizontal position
uint cstart = ustart - (BOXSIZE - 1) / 2;
uint cfinish = ufinish + (BOXSIZE - 1) / 2;
uint rstart = vstart - (BOXSIZE - 1) / 2;
uint i;
uint c, r;
for (c = cstart, i = 0; c < cfinish; c++, i++)
{
CSgxsq[i] = 0; CSgysq[i] = 0; CSgxgy[i] = 0;
for (r = rstart; r < rstart + BOXSIZE; r++)
{
gx = (image.image[c + 1, r] -
image.image[c - 1, r]) / 2.0f;
gy = (image.image[c, r + 1] -
image.image[c, r - 1]) / 2.0f;
CSgxsq[i] += gx * gx;
CSgysq[i] += gy * gy;
CSgxgy[i] += gx * gy;
}
}
// Now loop through u and v to form sums
evbest = 0;
for (v = vstart; v < vfinish; v++)
{
u = ustart;
// Form first sums for u = ustart
TSgxsq = 0.0f;
TSgysq = 0.0f;
TSgxgy = 0.0f;
for (i = 0; i < BOXSIZE; i++)
{
TSgxsq += CSgxsq[i];
TSgysq += CSgysq[i];
TSgxgy += CSgxgy[i];
}
for (u = ustart; u < ufinish; u++)
{
if (u != ustart)
{
// Subtract old column, add new one
TSgxsq += CSgxsq[u - ustart + BOXSIZE - 1] - CSgxsq[u - ustart - 1];
TSgysq += CSgysq[u - ustart + BOXSIZE - 1] - CSgysq[u - ustart - 1];
TSgxgy += CSgxgy[u - ustart + BOXSIZE - 1] - CSgxgy[u - ustart - 1];
}
find_eigenvalues(TSgxsq, TSgxgy, TSgysq, ref eval1, ref eval2);
// eval2 will be the smaller eigenvalue. Compare it with the one
// we've already got
if (eval2 > evbest)
{
ubest = u;
vbest = v;
evbest = eval2;
}
}
if (v != vfinish - 1)
{
// Update the column sums for the next v
for (c = cstart, i = 0; c < cfinish; c++, i++)
{
// Subtract the old top pixel
gx = (image.image[c + 1, v - (BOXSIZE - 1) / 2] -
image.image[c - 1, v - (BOXSIZE - 1) / 2]) / 2.0f;
gy = (image.image[c, v - (BOXSIZE - 1) / 2 + 1] -
image.image[c, v - (BOXSIZE - 1) / 2 - 1]) / 2.0f;
CSgxsq[i] -= gx * gx;
CSgysq[i] -= gy * gy;
CSgxgy[i] -= gx * gy;
// Add the new bottom pixel
gx = (image.image[c + 1, v + (BOXSIZE - 1) / 2 + 1] -
image.image[c - 1, v + (BOXSIZE - 1) / 2 + 1]) / 2.0f;
gy = (image.image[c, v + (BOXSIZE - 1) / 2 + 1 + 1] -
image.image[c, v + (BOXSIZE - 1) / 2 + 1 - 1]) / 2.0f;
CSgxsq[i] += gx * gx;
CSgysq[i] += gy * gy;
CSgxgy[i] += gx * gy;
}
}
}
}
/// <summary>
/// Version to find the best n patches within an image.
/// Not very efficiently implemented I'm afraid.
/// Now we expect the arguments ubest, vbest, evbest to be arrays of
/// size n.
/// </summary>
/// <param name="image">The image to scan</param>
/// <param name="n">Number of patches</param>
/// <param name="ubest">Array containing x-co-ordinates of n best patches</param>
/// <param name="vbest">Array containing y-co-ordinates of n best patches</param>
/// <param name="evbest">Array containing smallest eigenvalues of best patches (larger is better)</param>
/// <param name="BOXSIZE">The size of the patch to use</param>
/// <param name="ustart">The x-co-ordinate of the start of the search window</param>
/// <param name="vstart">The y-co-ordinate of the start of the search window</param>
/// <param name="ufinish">The x-co-ordinate of the end of the search window</param>
/// <param name="vfinish">The v-co-ordinate of the end of the search window</param>
public void find_best_n_patches_inside_region(
classimage_mono image,
uint n,
uint[] ubest, uint[] vbest,
float[] evbest,
uint BOXSIZE,
uint ustart, uint vstart,
uint ufinish, uint vfinish)
{
// Check that these limits aren't too close to the image edges.
// Note that we can't use the edge pixels because we can't form
// gradients here.
if (ustart < (BOXSIZE - 1) / 2 + 1)
{
ustart = (BOXSIZE - 1) / 2 + 1;
}
if (ufinish > image.width - (BOXSIZE - 1) / 2 - 1)
{
ufinish = (uint)(image.width - (BOXSIZE - 1) / 2 - 1);
}
if (vstart < (BOXSIZE - 1) / 2 + 1)
{
vstart = (BOXSIZE - 1) / 2 + 1;
}
if (vfinish > image.height - (BOXSIZE - 1) / 2 - 1)
{
vfinish = (uint)(image.height - (BOXSIZE - 1) / 2 - 1);
}
// Calculate the width we need to find gradients in.
uint calc_width = ufinish - ustart + BOXSIZE - 1;
// Arrays which store column sums of height BOXSIZE
float[] CSgxsq = new float[calc_width];
float[] CSgysq = new float[calc_width];
float[] CSgxgy = new float[calc_width];
// For the final sums at each position (u, v)
float TSgxsq = 0.0f, TSgysq = 0.0f, TSgxgy = 0.0f;
float gx, gy;
uint u = ustart, v = vstart;
float eval1 = 0, eval2 = 0;
// Initial stage: fill these sums for the first horizontal position
uint cstart = ustart - (BOXSIZE - 1) / 2;
uint cfinish = ufinish + (BOXSIZE - 1) / 2;
uint rstart = vstart - (BOXSIZE - 1) / 2;
uint i;
uint c, r;
for (c = cstart, i = 0; c < cfinish; c++, i++)
{
CSgxsq[i] = 0;
CSgysq[i] = 0;
CSgxgy[i] = 0;
for (r = rstart; r < rstart + BOXSIZE; r++)
{
gx = (image.image[c + 1, r] - image.image[c - 1, r]) / 2.0f;
gy = (image.image[c, r + 1] - image.image[c, r - 1]) / 2.0f;
CSgxsq[i] += gx * gx;
CSgysq[i] += gy * gy;
CSgxgy[i] += gx * gy;
}
}
float[,] evarray = new float[vfinish - vstart, ufinish - ustart];
// Now loop through u and v to form sums
for (v = vstart; v < vfinish; v++)
{
u = ustart;
// Form first sums for u = ustart
TSgxsq = 0.0f;
TSgysq = 0.0f;
TSgxgy = 0.0f;
for (i = 0; i < BOXSIZE; i++)
{
TSgxsq += CSgxsq[i];
TSgysq += CSgysq[i];
TSgxgy += CSgxgy[i];
}
for (u = ustart; u < ufinish; u++)
{
if (u != ustart)
{
/* Subtract old column, add new one */
TSgxsq += CSgxsq[u - ustart + BOXSIZE - 1] - CSgxsq[u - ustart - 1];
TSgysq += CSgysq[u - ustart + BOXSIZE - 1] - CSgysq[u - ustart - 1];
TSgxgy += CSgxgy[u - ustart + BOXSIZE - 1] - CSgxgy[u - ustart - 1];
}
find_eigenvalues(TSgxsq, TSgxgy, TSgysq, ref eval1, ref eval2);
// eval2 will be the smaller eigenvalue. Store in the array
evarray[v - vstart, u - ustart] = eval2;
}
if (v != vfinish - 1)
{
// Update the column sums for the next v
for (c = cstart, i = 0; c < cfinish; c++, i++)
{
// Subtract the old top pixel
gx = (image.image[c + 1, v - (BOXSIZE - 1) / 2]
- image.image[c - 1, v - (BOXSIZE - 1) / 2]) / 2.0f;
gy = (image.image[c, v - (BOXSIZE - 1) / 2 + 1]
- image.image[c, v - (BOXSIZE - 1) / 2 - 1]) / 2.0f;
CSgxsq[i] -= gx * gx;
CSgysq[i] -= gy * gy;
CSgxgy[i] -= gx * gy;
// Add the new bottom pixel
gx = (image.image[c + 1, v + (BOXSIZE - 1) / 2 + 1]
- image.image[c - 1, v + (BOXSIZE - 1) / 2 + 1]) / 2.0f;
gy = (image.image[c, v + (BOXSIZE - 1) / 2 + 1 + 1]
- image.image[c, v + (BOXSIZE - 1) / 2 + 1 - 1]) / 2.0f;
CSgxsq[i] += gx * gx;
CSgysq[i] += gy * gy;
CSgxgy[i] += gx * gy;
}
}
}
// Now: work out the best n patches which don't overlap each other
float best_so_far;
int xdist, ydist;
bool OKflag;
float next_highest = 1000000000000.0f;
for (i = 0; i < n; i++)
{
best_so_far = 0.0f;
for (uint y = 0; y < vfinish - vstart; y++)
{
for (uint x = 0; x < ufinish - ustart; x++)
{
if (evarray[y, x] > best_so_far && evarray[y, x] < next_highest)
{
// We've found a high one: check it doesn't overlap with higher ones
OKflag = true;
for (uint j = 0; j < i; j++)
{
xdist = (int)(x + ustart - ubest[j]);
ydist = (int)(y + vstart - vbest[j]);
xdist = (xdist >= 0 ? xdist : -xdist);
ydist = (ydist >= 0 ? ydist : -ydist);
if ((xdist < (int)BOXSIZE) && (ydist < (int)BOXSIZE))
{
OKflag = false;
break;
}
}
if (OKflag)
{
ubest[i] = x + ustart;
vbest[i] = y + vstart;
evbest[i] = evarray[y, x];
best_so_far = evarray[y, x];
}
}
}
}
next_highest = evbest[i];
}
}
/// <summary>
/// Do a search for patch in image within an elliptical region. The
/// search region is parameterised an inverse covariance matrix (a distance of
/// NO_SIGMA is used). The co-ordinates returned are those of centre of the patch.
/// </summary>
/// <param name="image">The image to search</param>
/// <param name="patch">The patch to search for</param>
/// <param name="centre">The centre of the search ellipse</param>
/// <param name="PuInv">The inverse covariance matrix to use</param>
/// <param name="u">The x-co-ordinate of the best patch location</param>
/// <param name="v">The y-co-ordinate of the best patch location</param>
/// <param name="uBOXSIZE">The size of the image patch (TODO: Shouldn't this be the same as the size of patch?)</param>
/// <returns>true if the a good match is found (above CORRTHRESH2), false otherwise</returns>
public static bool elliptical_search(
classimage_mono image,
classimage_mono patch,
Vector centre, //1 dimensional x,y coords
MatrixFixed PuInv, //2x2 matrix
ref uint u, ref uint v,
Vector vz,
uint uBOXSIZE,
classimage_mono outputimage,
bool show_ellipses,
bool calibrating,
Random rnd)
{
float aspect;
if ((vz[1] != 0) && (vz[0] != 0))
{
aspect = Math.Abs(vz[0] / vz[1]);
if (aspect < 0.5)
{
aspect = 0.5f;
}
if (aspect > 1.5)
{
aspect = 1.5f;
}
}
else
{
aspect = 1;
}
int vz_x = 0; // (int)(vz[0] / 2);
int vz_y = 0; // (int)(vz[1] / 2);
// We want to pass BOXSIZE as an unsigned int since it is,
// but if we use it in the if statements below then C++ casts the
// whole calculation to unsigned ints, so it is never < 0!
// Force it to get the right answer by using an int version of BOXSIZE
int BOXSIZE = (int)uBOXSIZE;
// The dimensions of the bounding box of the ellipse we want to search in
uint halfwidth = (uint)(NO_SIGMA * aspect /
Math.Sqrt(PuInv[0, 0] - PuInv[0, 1] * PuInv[0, 1] / PuInv[1, 1]));
uint halfheight = (uint)(NO_SIGMA / aspect /
Math.Sqrt(PuInv[1, 1] - PuInv[0, 1] * PuInv[0, 1] / PuInv[0, 0]));
// stop the search ellipse from expanding to large sizes!
uint MAX_SEARCH_RADIUS = uBOXSIZE * 2;
if (halfwidth > MAX_SEARCH_RADIUS)
{
halfwidth = MAX_SEARCH_RADIUS;
}
if (halfheight > MAX_SEARCH_RADIUS)
{
halfheight = MAX_SEARCH_RADIUS;
}
// don't allow the ellipse to get too small
uint MIN_SEARCH_RADIUS = uBOXSIZE / 2;
if (halfwidth < MIN_SEARCH_RADIUS)
{
halfwidth = MIN_SEARCH_RADIUS;
}
if (halfheight < MIN_SEARCH_RADIUS)
{
halfheight = MIN_SEARCH_RADIUS;
}
int ucentre = (int)(centre[0] + 0.5);
int vcentre = (int)(centre[1] + 0.5);
// Limits of search
int urelstart = -(int)(halfwidth);
int urelfinish = (int)(halfwidth);
int vrelstart = -(int)(halfheight);
int vrelfinish = (int)(halfheight);
// Check these limits aren't outside the image
if (ucentre + urelstart + vz_x - (BOXSIZE - 1) / 2 < 0)
{
urelstart = (BOXSIZE - 1) / 2 - ucentre - vz_x;
}
if (ucentre + urelfinish + vz_x - (BOXSIZE - 1) / 2 > (int)(image.width) - BOXSIZE)
{
urelfinish = image.width - BOXSIZE - ucentre - vz_x + (BOXSIZE - 1) / 2;
}
if (vcentre + vrelstart + vz_y - (BOXSIZE - 1) / 2 < 0)
{
vrelstart = (BOXSIZE - 1) / 2 - vcentre - vz_y;
}
if (vcentre + vrelfinish + vz_y - (BOXSIZE - 1) / 2 > (int)(image.height) - BOXSIZE)
{
vrelfinish = (int)(image.height) - BOXSIZE - vcentre - vz_y + (BOXSIZE - 1) / 2;
}
// Search counters
int urel, vrel;
float corrmax = 1000000.0f;
float corr;
// For passing to and_correlate2_warning
float sdpatch = 0, sdimage = 0;
// Do the search
float max_dist = MAX_SEARCH_RADIUS;
float v1 = PuInv[0, 0];
float v2 = PuInv[0, 1];
float v3 = PuInv[1, 1];
float urelsq, vrelsq;
bool inside_ellipse;
int xx, yy, xx2, yy2;
for (urel = urelstart; urel <= urelfinish; urel += 1)
{
urelsq = urel * urel;
float urel2 = v1 * urelsq;
inside_ellipse = false;
for (vrel = vrelstart; vrel <= vrelfinish; vrel += 1)
{
vrelsq = vrel * vrel;
if (urel2
+ 2 * v2 * urel * vrel
+ v3 * vrelsq < NO_SIGMA * NO_SIGMA)
{
if ((show_ellipses) || (calibrating))
{
if (!inside_ellipse)
{
xx = ucentre + urel + vz_x - (BOXSIZE - 1) / 2;
yy = vcentre + vrel + vz_y - (BOXSIZE - 1) / 2;
xx2 = xx * outputimage.width / image.width;
yy2 = yy * outputimage.height / image.height;
outputimage.image[xx2, yy2] = (Byte)255;
}
inside_ellipse = true;
}
// searching within the ellipse is probablistic,
// using something like a gaussian distribution
float dist = (float)Math.Sqrt(urelsq + vrelsq) / max_dist;
if (rnd.Next(1000) < search_probability(dist))
{
int offset_x = ucentre + urel + vz_x;
int offset_y = vcentre + vrel + vz_y;
corr = correlate2_warning(0, 0, BOXSIZE, BOXSIZE,
offset_x - (BOXSIZE - 1) / 2,
offset_y - (BOXSIZE - 1) / 2,
ref patch, ref image, ref sdpatch, ref sdimage);
if (corr <= corrmax)
{
if (sdpatch < CORRELATION_SIGMA_THRESHOLD)
{
// cout << "Low patch sigma." << endl;
}
else
if (sdimage < CORRELATION_SIGMA_THRESHOLD)
{
// cout << "Low image sigma." << endl;
}
else
{
corrmax = corr;
u = (uint)offset_x;
v = (uint)offset_y;
}
}
//show ellipses in the output image
/*
* if ((show_ellipses) || (calibrating))
* {
* int xx = offset_x;
* int yy = offset_y;
* int xx2 = xx * outputimage.width / image.width;
* int yy2 = yy * outputimage.height / image.height;
*
* if (!calibrating)
* {
* outputimage.image[xx2, yy2, 2] = (Byte)255;
* }
* else
* {
* outputimage.image[xx2, yy2, 0] = (Byte)255;
* outputimage.image[xx2, yy2, 1] = (Byte)255;
* outputimage.image[xx2, yy2, 2] = 0;
* }
* }
*/
}
}
else
{
if ((show_ellipses) || (calibrating))
{
if (inside_ellipse)
{
xx = ucentre + urel + vz_x;
yy = vcentre + vrel + vz_y;
xx2 = xx * outputimage.width / image.width;
yy2 = yy * outputimage.height / image.height;
outputimage.image[xx2, yy2] = (Byte)255;
}
inside_ellipse = false;
}
}
}
}
if (corrmax > CORRTHRESH2)
{
return(false);
}
return(true);
}
/**************************Initialise Known Features**************************/
/// <summary>
/// Initialise the Scene_Single class with some known features, read from the
/// Settings. Each known feature has its own section, starting with
/// <code>[KnownFeature1]</code> and counting upwards. The feature type is
/// identified with the entry <code>FeatureMeasurementModel=</code>. Further
/// settings are loaded by the feature measurement model itself.
/// </summary>
/// <param name="model_creator"></param>
/// <param name="sim_or_rob"></param>
/// <param name="scene"></param>
/// <returns></returns>
public static uint initialise_known_features(Settings settings,
Feature_Measurement_Model_Creator model_creator,
Sim_Or_Rob sim_or_rob,
Scene_Single scene,
String path,
float MAXIMUM_ANGLE_DIFFERENCE)
{
uint feature_no = 1;
uint num_features = 0;
Settings.Section section = null;
do
{
// Step through the section names
String section_name = "KnownFeature" + Convert.ToString(feature_no);
section = settings.get_section(section_name);
// Does this section exist?
if (section == null)
{
return(num_features);
}
ArrayList values = section.get_entry("FeatureMeasurementModel");
if (values == null)
{
Debug.WriteLine("No FeatureMeasurementModel entry under the section [" +
section_name + "] in initalisation file.");
}
else
{
String type = (String)values[0];
Feature_Measurement_Model f_m_m =
model_creator.create_model(type, scene.get_motion_model(), MAXIMUM_ANGLE_DIFFERENCE);
if (f_m_m == null)
{
Debug.WriteLine("Unable to create a feature measurement model of type " +
type + " as requested in initalisation file.");
}
else
{
// Initialise the feature measurement model with any settings
f_m_m.read_parameters(settings);
// Read the feature state
Vector yi = new Vector(3);
Vector xp_orig = new Vector(7);
f_m_m.read_initial_state(section, yi, xp_orig);
// Initialise the feature
classimage_mono identifier =
sim_or_rob.initialise_known_feature(f_m_m, yi, section, path);
if (identifier == null)
{
Debug.WriteLine("Trouble reading known feature " +
section_name + " : skipping.");
}
else
{
scene.add_new_known_feature(identifier, yi, xp_orig, f_m_m, feature_no);
Debug.WriteLine("Added known feature " + Convert.ToString(feature_no));
num_features++;
}
}
}
feature_no++;
}while (section != null);
return(num_features);
}
//**************************Write Image Patch to Disk**************************
/// <summary>
/// Save the currently-selected patch to a file.
/// </summary>
/// <param name="name"></param>
public void write_patch(String name)
{
classimage_mono hip = new classimage_mono();
hip.createImage((int)Camera_Constants.BOXSIZE, (int)Camera_Constants.BOXSIZE);
if (location_selected_flag)
{
// Copy the selected patch to the save space patch
copy_into_patch(image, hip, uu, vv);
hip.SaveAsBitmapMono(name);
}
}
/// <summary>
/// Constructor for partially-initialised features.
/// </summary>
/// <param name="id">reference to the feature</param>
/// <param name="lab"></param>
/// <param name="list_pos"></param>
/// <param name="scene"></param>
/// <param name="h"></param>
/// <param name="p_i_f_m_m"></param>
public Feature(classimage_mono id, uint lab, uint list_pos,
Scene_Single scene, Vector h,
Partially_Initialised_Feature_Measurement_Model p_i_f_m_m,
Vector feature_colour)
{
feature_measurement_model = p_i_f_m_m;
partially_initialised_feature_measurement_model = p_i_f_m_m;
fully_initialised_feature_measurement_model = null;
// Stuff below substituted from Feature_common
// Feature_common(id, lab, list_pos, scene, h);
feature_constructor_bookeeping();
identifier = id;
label = lab;
position_in_list = list_pos; // Position of new feature in list
position_in_total_state_vector = 0; // This should be set properly
colour = feature_colour;
//when feature is added
// Save the vehicle position where this feature was acquired
scene.get_motion_model().func_xp(scene.get_xv());
//xp_orig = scene.get_motion_model().get_xpRES();
xp_orig = new Vector(scene.get_motion_model().get_xpRES());
// Call model functions to calculate feature state, measurement noise
// and associated Jacobians. Results are stored in RES matrices
// First calculate "position state" and Jacobian
scene.get_motion_model().func_xp(scene.get_xv());
scene.get_motion_model().func_dxp_by_dxv(scene.get_xv());
// Now ask the model to initialise the state vector and calculate Jacobians
// so that I can go and calculate the covariance matrices
partially_initialised_feature_measurement_model.func_ypi_and_dypi_by_dxp_and_dypi_by_dhi_and_Ri(h, scene.get_motion_model().get_xpRES());
// State y
//y = partially_initialised_feature_measurement_model.get_ypiRES();
y = new Vector(partially_initialised_feature_measurement_model.get_ypiRES());
// Temp_FS1 will store dypi_by_dxv
MatrixFixed Temp_FS1 =
partially_initialised_feature_measurement_model.get_dypi_by_dxpRES() *
scene.get_motion_model().get_dxp_by_dxvRES();
// Pxy
Pxy = scene.get_Pxx() * Temp_FS1.Transpose();
// Pyy
Pyy = Temp_FS1 * scene.get_Pxx() * Temp_FS1.Transpose()
+ partially_initialised_feature_measurement_model.get_dypi_by_dhiRES()
* partially_initialised_feature_measurement_model.get_RiRES()
* partially_initialised_feature_measurement_model.get_dypi_by_dhiRES().Transpose();
// Covariances of this feature with others
int j = 0;
foreach (Feature it in scene.get_feature_list_noconst())
{
if (j < position_in_list)
{
// new Pypiyj = dypi_by_dxv . Pxyj
// Size of this is FEATURE_STATE_SIZE(new) by FEATURE_STATE_SIZE(old)
MatrixFixed m = it.get_Pxy();
MatrixFixed newPyjypi_to_store = (Temp_FS1 * m).Transpose();
//add to the list
matrix_block_list.Add(newPyjypi_to_store);
}
j++;
}
known_feature_label = -1;
}
